Dispatch

Dispatch is a non-custodial desk of trading agents for following verified on-chain traders on Base. They continuously read the full ledger, rebuild trading histories for active wallets, and admit a trader to the roster only after their results prove consistent across independent periods, beat chance and hold up under real execution delay and market depth.

When a rostered trader acts, the agents prepare, price, simulate and settle the corresponding action for each client account within seconds, behind a sequence of safety guards that reject untradeable assets, price chasing and excessive concentration. Client funds remain in smart accounts owned by the clients' own wallets. Signed on-chain permissions allow trading and capped fees but never transfers to third parties.

The $DSP token aligns holders with the agents. Holding it lowers performance fees, unlocks larger trader desks and will give priority access to scarce capacity on the most sought-after traders. This document sets out the problem, the architecture, the security model, the token's role and the roadmap.

Public blockchains are the most transparent markets ever built. Every order, every position and every exit is written to a ledger that anyone can read. In principle this means the best decisions in the market are visible to everyone. In practice almost nobody can use them. The record is enormous and noisy, most visible activity is automated or adversarial, and the few wallets with genuine, repeatable judgment are hidden among hundreds of thousands of others.

Even when such a wallet is found, acting on its moves is hard. Markets on Base move in seconds. Fresh tokens can be built to trap buyers. Liquidity is thin and changes block by block. A person who sees a good trade and repeats it by hand is usually too late, too large for the market or exposed to a token that cannot be sold back.

Dispatch was designed to remove each of these obstacles in turn. The agents treat the ledger as raw material and turn it into a short, continuously verified roster of traders worth following. They measure whether each trader's edge survives real execution before a single client unit is committed. When a rostered trader acts, the agents size the move to each account, check it against a sequence of safety guards and settle it within seconds. Throughout, client funds remain in client-owned smart accounts. The agents can trade on a client's behalf, but they cannot take custody of the money.

Base has become one of the busiest networks for on-chain trading. Low fees and fast blocks have drawn a constant flow of new markets: token launches, community projects, AI-agent tokens and a growing set of established assets with deep liquidity. The same conditions that make the network attractive make it hard to navigate. New markets appear every minute, most of them fail, and the ones that succeed often do so within hours.

In this environment the scarce asset is information. Anyone can buy any token. Very few people can tell, at the moment it matters, which buyers know what they are doing. That knowledge already exists in the ledger, written by the traders themselves. Agents that can extract it reliably and act on it safely give ordinary participants something they could not build alone: the judgment of the best traders on the network, applied to their own capital, at their own risk level, without handing that capital to anyone.

Dispatch is a desk of agents that combines what each of these lacks: machine speed without custody, a ranking built on durable, followable results rather than headline numbers, and the same safety discipline on every action, whoever the client is and whatever the size of their account.

The agents are organised in five layers with clear boundaries. Information flows from left to right: raw chain data becomes candidate signals, candidates become a verified roster, roster activity becomes proposed actions, and proposed actions become settled trades inside client accounts. Results flow back from right to left. Every settled trade, every skipped action and every outcome is fed back to the Signal layer, so the roster is judged on how it performs for clients and not only on history.

Each layer can be improved, tested and scaled independently. This separation is deliberate. Discovery research can run at full depth without touching the live trading path. Risk rules can be tightened without rewriting execution. Most importantly, the Account layer sits outside the reach of every other layer: nothing upstream holds the authority to move client funds anywhere except back to the client.

Everything starts with a complete, faithful picture of what happens on the chain. The Signal layer reads Base block by block and rebuilds the economic meaning of each transaction. A swap is recognised as a swap whatever venue or route carried it. A transfer between two wallets of the same owner is told apart from a sale. A purchase paid in a stablecoin or another token is valued consistently with one paid in ETH.

On top of this event stream the layer reconstructs full trading histories for wallets of interest: every entry and exit, the size of each position over time, and the value of what is still held. Fresh markets often have no reliable external price. For these the layer derives prices directly from on-chain pool state, so a token launched an hour ago can be valued as accurately as an established one.

The layer is built for completeness. Discovery begins from the markets themselves: every newly created pool on the network, rather than lists published by aggregators, which see only a small fraction of real activity. In our measurements, aggregator coverage represented a few hundred tokens over a period in which hundreds of thousands of new markets were opened on Base.

Finding a trader worth following is a search for a few true signals in a very large haystack. Each stage below narrows the field on evidence, never on reputation, follower count or self-reported results. The volumes are from recent discovery cycles on Base.

The steepest cut is the last one. Many wallets look excellent on paper and still cannot be followed: their edge depends on being first in the block, on a position size the market cannot absorb twice, or on assets that cannot be sold back. Execution testing exists to find these cases before a client does. An admission rate of a few per hundred thousand is what remains when all of these effects are removed.

Market scan over 90 days; wallet stages over the latest 30-day cycle. Cycles run continuously, and the admitted roster is reviewed against live results.

Admission is a sequence of independent tests. Each one closes a specific way in which past results can mislead, and a candidate must pass all of them. No single score can compensate for a failed test.

Admission is not permanent. Once rostered, a trader is measured on the results they produce for clients. A trader whose live behaviour drifts from the profile that earned the seat loses weight or is removed. Traders may also change wallets or move holdings between their own addresses. The agents recognise this and follow the position to its real exit rather than losing track of it.

When a rostered trader acts, the agents have a few seconds to respond while the opportunity is still close to the price the trader saw. The execution path is organised around this budget. New blocks are processed as they arrive. The trader's action is recognised and classified, and the corresponding action for each following account is prepared in parallel.

Each prepared action is priced against all available liquidity, and the best route is chosen for the exact size of that account. The action is simulated against the current state of the network before it is sent. The transaction carries a strict minimum outcome, so it either settles within a narrow band around the simulated price or does not settle at all. When an action does not settle because the network moved first, the agents re-evaluate it from scratch under the same guards before deciding whether to try again.

Exits follow the trader just as closely as entries. When a trader sells part of a position, each following account sells the same share of its own position. When a position reaches a predefined multiple of its cost, the original stake is returned to the account and the remainder continues to ride, so only the profit stays exposed.

Every action passes the same sequence of guards before funds move. The guards are ordered from cheapest to most thorough, so most unsafe actions are rejected immediately and only promising ones reach full simulation. A failed guard skips the action and records why. Nothing is forced through, and every decision can be traced afterwards.

Dispatch is built so that the operator never holds client funds. Each client receives a dedicated smart account on Base, owned by the client's own wallet. Deposits go into that account and stay there.

To let the agents act, the client signs a small set of narrowly scoped permissions. Each permission is enforced on-chain, not by policy, and allows exactly one kind of action. One permits trading through approved liquidity routes. One permits returning funds to the owner's wallet. One permits a capped performance fee. Each has its own limits and expiry. There is no permission to send funds to an arbitrary address, so even in the worst case funds can only be traded inside the account or returned to its owner.

Custody by design decides where the money lives. This sheet describes what protects it while the agents trade, and what happens when something goes wrong. The principle is the same throughout: the controls that matter are enforced on chain, where a client can verify them, and the operational layer around them is built to detect, bound and recover.

The gate contract

Every trade the agents make passes through GuardedSwap, a verified contract that sits between the client account and the liquidity aggregator. The agents' trading permission points at the gate and at nothing else. The gate decodes the route it is handed and enforces its own rules before any funds move, regardless of who signed the transaction:

The gate has no owner, no upgrade path and holds no funds between calls. Its source is published and verified against the deployed bytecode: deployment on Base · verified source (Sourcify). The current deployment is also linked from the app.

What the gate does and does not promise

The gate limits damage; it does not remove risk. If the agents' trading key were ever compromised, the attacker could still trade badly inside the account, and the worst case is bounded by the loss budget and the hourly limits rather than by the size of the account. That bound is the design goal: a known, limited loss instead of an open-ended one. Market risk, token risk and the risk of defects in the surrounding software remain, as described under Risk factors.

Procedures

Numeric limits, alert channels and key custody details are operational parameters and are intentionally not published here; the contract rules above are the ones a client can verify independently.

Agents that trade on behalf of others have to keep working when their surroundings do not. The runtime reads the network through several independent providers. When one of them slows down or refuses a request, the next one takes over without waiting. Processing state is written to disk continuously, so a restart resumes from the exact block where work stopped and never replays or skips an action.

Independent watchdogs check the agents from the outside. They confirm that the agents are current with the chain, that every trader action received a response, and that the accounts they manage hold what the records say they should. Anything unusual raises an alert to the operators within minutes. Every action, skip and recovery is logged with its reason, so any outcome a client sees can be traced to the decision that produced it.

Most clients do not want to assemble a roster themselves. Ready-made strategies package the verified roster into desks with a defined character: which traders are followed, how much of the account each entry uses, when the original stake is returned, and what fee applies. Clients can also add up to five wallets of their own choice next to a strategy.

Every account shows its own performance: return in percent and in ETH, an annualised rate before and after fees, and a chart of the account's value since its first deposit. The figures are calculated from the account itself and not from a model or a backtest.

$DSP is the access and alignment layer of Dispatch. Its utility is tied directly to the product. Every benefit corresponds to something the agents already do or are building, and none depends on speculation alone.

Supply, allocation and distribution details are published separately by the token issuer. This blueprint covers only how the token interacts with the product.

Holding $DSP lowers the fee on profit, step by step. The schedule keeps strategies separate because they carry different returns and different costs to run, and it never raises the fee for anyone above today's base rate. The wallet balance is read at sign-in, with nothing to stake, lock or migrate.

The core agents are live and trading on Base. The next releases build a holder layer on top of it and then add features that matter at larger scale, where demand for the best traders starts to exceed what their markets can absorb.

1. 01Live now

   Core agents

   • Three ready-made strategies
   • Follow up to five wallets of your choice
   • Performance view with return and APR
   • Non-custodial smart accounts on Base
2. 02Next release

   Holder layer

   • $DSP fee tiers
   • Larger trader desks for holders
   • Wallet-following as a holder feature
3. 03Following releases

   Scale layer

   • Priority capacity on the most popular traders
   • Fees payable in $DSP at a discount
   • Real-time trade signals for holders

Trading digital assets carries substantial risk, and following other traders does not remove it. Past results of any trader, however carefully verified, do not guarantee future results. Markets on Base can move sharply within minutes, newly launched tokens may lose most or all of their value, and liquidity can disappear. Execution happens after the trader's own action, so prices can differ from theirs. Smart accounts and the underlying contracts are software and may contain undiscovered faults. Network congestion or provider outages can delay actions. Clients should deposit only what they can afford to lose and should choose a strategy that matches their tolerance for loss.

The best decisions on Base are already public. What has been missing is a way to find the few that are real, prove they can be followed, and act on them safely without giving up custody. Dispatch is built around that gap: chain-wide discovery, evidence-based admission, guarded execution and client-owned accounts, operating continuously as one desk of agents. The holder layer and the scale features on the roadmap extend the same principles as the network and the user base grow.

This document is for information only. It is not an offer of securities or investment advice, and it does not create any obligation on the part of Dispatch. Features marked as planned may change.